From what I understand, normalizing wav files before encoding to mp3 produces better results and there are different programs for different formats. The drawbacks seem to be that songs intended to be softer gain too much volume when brought up to the levels of a louder file. There is also the issue with clipping, of which I am currently reading up on as I don't really understand it.
I've heard people gripe that peaks can also be cut from songs while normalizing, although it doesn't make sense to me if all files are matched to the loudest registered db.
Anyway, I was just wondering if this was really a big issue and if someone would like to expound on the whole normalizing thing.
Normalizing is simply the process of taking the highest trancesent peak in the file and raising that to the maximum possible 16-bit value and then raising the rest of the file with a logarithmic value based on that one peak wave.
Clipping is simply impossible when you normalize because there's no peaks being greater then the maximum possible 16-bit value. If you manually raise the volume of the WAV file with a linear value (ie- +6dB) there's a good chance you may end up clipping the file.
Peaks are cut when people start using dynamic compression to iron out all the peaks before they boost the volume manually or normalize. Or if they normalize using RMS values and not the peak value they'll definitely be clipping a few of the occasional spikes.